If I have an expression and it involves computations with extremely large numbers. Double precision is bound to fail in my case.
And when I used Matlab 2007a's symbolic toolbox, which I believe uses Maple 10 or Maple 9's engine, it supposedly should always return correct value, because its symbolic, with infinite precision. But it still fails for some parameters which makes the numbers too large.
Now if I compute the same expression in Maple 11, it is even worse. The calculation fails for a lot more parameters. The failing parameter set for previous Maple version is now a subset of the failing parameter set for the Maple 11. The difference is huge -- Maple 11 gives wrong values almost all the time, even for very friendly parameters.
What might be the cause of this problem? Is "Digits:=1000" the only cure? I've tried changing Digits to 10000, it still failed on almost all parameters the previously worked in older version of Maple... (via Matlab 2006a and 2007a, I've tested)...